This invention relates generally to gas turbine engines, and more particularly, to methods for determining sensor locations in gas turbine engines.
Optimal sensor placement facilitates accurately determining the average value of an operating parameter, such as the temperature of a gas path airflow. Generally, there are two known methods used to select the circumferential location of sensors positioned downstream of a gas turbine engine combustor. A first known method randomly distributes sensors circumferentially as widely as possible based on the clear space available on the engine or other criteria. A second known method uses engine testing or burner rig tests to simulate gas turbine engine combustor performance. A very large number of data points throughout the operating regime are recorded and analyzed to determine a set of sensor locations that improves temperature measurements. The empirical data is used to determine an acceptable measure of the average bulk temperature of the gas flow path. However, performing the burner rig tests, and recording and analyzing the data may be an expensive and time consuming process.
These sensors are generally used together to measure an operating parameter, such as exhaust temperature, by calculating an average value, a maximum value, a minimum value, or other characteristic of the given operating parameter. Should one or more of the sensors fail, the average value of the parameter calculated using the remaining sensors may be corrupted. Generally, when a faulty sensor is detected, it is removed from the average calculation and is replaced at the next opportunity.
Removing a faulty sensor from the average calculation reduces the accuracy of the average calculation of the remaining sensors. Consequently, at least some known engines include supplemental sensors so that the average measurement is minimally impacted when a sensor fails. The degree to which the system measurement is impacted is proportional to the number of remaining functional sensors, i.e., the more functional sensors remaining, the less impact to the system. Moreover, using a large number of sensors to reduce the system impact of a fault increases costs and sensor system complexity, inherently increases the failure rate of sensors, and may lead to increased measurement error.
Other known detection methods do not compensate for a faulty sensor, but rather compute the average using n−1 sensors. Alternately, other known detection methods may replace the sensor value with a recent historical value. Failing sensors introduce a more complex compensation process because of the difficulty in determining whether a sensor has started failing, or is drifting. As the sensor begins to drift, the averaging process attenuates its impact.